AI From The Trenches

In this episode, JD talks with Kris (Pyramidal) about building a time-series foundation model trained on millions of hours of EEG, why healthcare AI is defensible but slow, and why regulation, reimbursement, and incentive misalignment are the real bottlenecks.
The Bet: A foundation model for EEG brainwaves that makes brain disorder diagnosis scalable, starting with epilepsy and ICU monitoring.
We cover Kris’s path (finance → Spotify audio wave detection → Google healthcare → founding Pyramidal), how they chose the initial wedge, what “unsupervised” training looks like for brain activity, why in-person selling is back, and what “human-computer symbiosis” could look like longer-term (implants, consumer neurotech, multi-modal models).
Takeaway: If you can clear regulation + reimbursement, the tech becomes defensible and the market becomes enormous.

How do you stay on top of science when over 500 cancer papers are published every day?In this episode of AI From the Trenches, Duncan talks with Dr Oliver Bogler about AI in cancer research, scientific curiosity, and what happens when the literature becomes impossible for humans to track on their own.They cover how AI can help with literature surveillance, what might replace the traditional scientific paper, what Humane’s AI Pin got right and wrong, and why synthetic biology could be one of the most powerful uses of AI in the coming years.In this conversation:

• Why most scientists are not early adopters of new techwhat it felt like when the literature went from manageable to impossible
• How AI could rewrite how we search, read and publish science
• Lessons from building NanCi together at the National Cancer Institute
• Humane AI Pin, wearables and voice native AI agents
• AlphaFold, synthetic biology and the future of drug discovery

JD and Duncan reflect on their journey from IBM Watson to founding Barnacle Labs.

They talk about building AI too early, why Watson struggled to become a real product, what went wrong with Humanise, and how early access to GPT-3 reshaped their thinking.

The conversation covers AI hype, agents, biotech, client responsibility, and what it actually takes to build systems that matter.

A practical, honest discussion about AI beyond the buzzwords.

Shopify’s Rich Brown tells the inside story of Scout, the internal AI tool that started as a side project, spread through the company, hit leadership meetings and even reached the board.

This episode cuts through the hype and shows how AI actually gets adopted in a large organisation:

• How non-engineers used tools like Cursor and Claude Code to build real products
• How Scout merged support, sales and community feedback into a single intelligence layer
• How video pain points became searchable with Scout TV
• How Shopify’s “kill the toil” culture created the conditions for rapid AI adoption
• Why the old rules of product still matterIf you work in product, support, AI strategy or enterprise transformation, this is the most honest look you’ll get at AI in the trenches.

If you work in product, support, AI strategy or enterprise transformation, this is the most honest look you’ll get at AI in the trenches.

Muna started in old-school advertising, making one of the UK’s best-performing TV ads. Today, she runs Circler, an AI-driven platform for high-end secondhand fashion that personalises outfits down to body shape, style and budget.

In this episode of AI From The Trenches, she explains how a “non-technical” founder built a serious AI product, automated styling with agents, and turned WhatsApp into a 7 percent conversion channel.

We get into:

• Why secondhand fashion is a data problem first (size, style, price)
• How Muna went from manual tagging and human stylists to AI-tagged inventory
• Building with a tiny tech team that “runs on AI agents”
• The WhatsApp experiment that crushed ecommerce benchmarks
• What marketing directors get wrong about “plug and play” AI
• Why you don’t need to be an AI expert – but you do need the right guideIf you’re a marketer, founder or operator trying to actually use AI in a business, this is a very honest, practical case study from someone who’s done it.